I am working with a data sets to compare two clinical tests that yield either a positive result (1) or negative result (0). The clinical tests use EEG responses, so as the testing time increases, the accuracy tends to increase as well. Given the population that we are working with, we are wanting to determine a general expected sensitivity of the test as a function of the time spent testing.
My current implementation involves taking a group of individuals that have the disease, and assessing the accuracy of the test over time (taken at each minute of testing), so in R, this looks like:
glmer(test_result ~ clinical_test * testing_Time + (1|Individual), family="binomial")
where test_result is the binary outcome, clinical_test is categorical, and testing_Time is a continuous variable.
Would the probability (resulting from the corresponding logit found using glmer) of correct classification be equivalent to the expected sensitivity of the test at n minutes? Or is there a better way to do this?